DESCRIPTION (adapted from the Abstract): The principal aim of this application is the further development of new methods for analyzing observational data bases and randomized trials of HIV-infected persons. The proposed approaches are based either on (1) the estimation of new classes of causal models, or (2) new methods for analyzing semi- or non-parametric models in the presence of both informative and non-informative missing data. The new classes of causal models include structural nested models, marginal structural models, direct effect structural nested models, and continuous time structural nested models. Many of the new methods are fundamentally "epidemiologic" in that they require data on time-dependent confounding factors, that is, risk factors for outcomes that also predict subsequent treatment with the drug or co-factor under study. The proposed methods of analysis will improve upon previous methods in the following ways: First, the new methods are the best methods available to estimate the effect o a treatment (e.g., AZT) or a co-factor (e.g., marijuana) on an outcome of interest (e.g., time to AIDS or HIV RNA levels) from observational data, when symptoms of HIV disease (e.g., thrush, fever) are simultaneously confounders and intermediate variables. The researchers will use the new methods to estimate treatment and co-factor effects on the evolution of CD4 counts and on time to progression of HIV-disease among subjects in the Multicenter AIDS Cohort Study (MACS). Results will be compared with results obtained using standard methods. Second, the new methods are the best methods available to adjust for dependent censoring, non-random non-compliance, treatment cross-over or termination, and the concurrent effect of additional non-randomized treatments in randomized clinical trials. For example, in ACTG Trial 002 of the effect of high-dose versus low-dose AZT on the survival of AIDS patients, patients in the low-dose arm took more aerosolized pentamidine (a non-randomized treatment). The new methods are the best methods available to incorporate information efficiently on surrogate markers (e.g., HIV RNA) in order to stop, at the earliest possibl moment, randomized trials of the effect of a treatment on a survival time outcome (e.g., time to AIDS). Specifically, the researchers will construct a valid a-level test of the null hypothesis of no effect of treatment on surviva that incorporates data on the evolution of multivariate surrogate markers (e.g., CD4-count and HIV RNA), with power exceeding that of the log rank test. They shall use the new methods to analyze further ACTG Trial 002 as well as ACTG Trials 021 and 175. These last two trials are comparisons of the effects of various chemotherapeutic agents on the opportunistic infection and survival experience of HIV-infected patients.